sparc64: Fix regression in non-hypervisor TLB flush xcall

Previously, %g2 would end up with the value PAGE_SIZE, but after the
commit mentioned below it ends up with the value 1 due to being reused
for a different purpose. We need it to be PAGE_SIZE as we use it to step
through pages in our demap loop, otherwise we set different flags in the
low 12 bits of the address written to, thereby doing things other than a
nucleus page flush.

Fixes: a74ad5e660 ("sparc64: Handle extremely large kernel TLB range flushes more gracefully.")
Reported-by: Meelis Roos <mroos@linux.ee>
Tested-by: Meelis Roos <mroos@linux.ee>
Signed-off-by: James Clarke <jrtc27@jrtc27.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit is contained in:
James Clarke 2019-05-29 22:31:31 +01:00 committed by David S. Miller
parent f2c7c76c5d
commit d3c976c14a

View File

@ -587,7 +587,7 @@ xcall_flush_tlb_kernel_range: /* 44 insns */
sub %g7, %g1, %g3
srlx %g3, 18, %g2
brnz,pn %g2, 2f
add %g2, 1, %g2
sethi %hi(PAGE_SIZE), %g2
sub %g3, %g2, %g3
or %g1, 0x20, %g1 ! Nucleus
1: stxa %g0, [%g1 + %g3] ASI_DMMU_DEMAP
@ -751,7 +751,7 @@ __cheetah_xcall_flush_tlb_kernel_range: /* 44 insns */
sub %g7, %g1, %g3
srlx %g3, 18, %g2
brnz,pn %g2, 2f
add %g2, 1, %g2
sethi %hi(PAGE_SIZE), %g2
sub %g3, %g2, %g3
or %g1, 0x20, %g1 ! Nucleus
1: stxa %g0, [%g1 + %g3] ASI_DMMU_DEMAP